Online shopping has made buying and returning clothing seamless for consumers. But for retailers, returns have become an expensive problem: The process requires retailers to pay to retrieve any garment that does not fit before inspecting it and deciding what to do with it, often at a loss. Virtual try-ons powered by generative artificial intelligence are emerging as the most commercially viable fix.
Platform companies, retailers, and a growing number of startups are now embedding generative AI virtual try-on technology directly into product pages, search results, and checkout flows.
According to the U.S. National Retail Federation, 15.8% of annual retail sales were returned in 2025, totaling $849.9 billion, as reported by CNBC.
For online sales, that figure rose to 19.3%, according to the same NRF data cited by CNBC. Gen Z shoppers aged 18 to 30 averaged nearly eight online returns per person last year, the NRF found.
Returns Eat Directly Into Margins
Uncertainty over fit is the primary driver of both returns and abandoned shopping carts, Ed Voyce, founder and CEO of AI startup Catches, told CNBC. Most returned items never make it back to shelves and often cost retailers more to process than the value of the refund. NRF data shows that 82% of consumers consider free returns essential, yet the cost of providing them is becoming unsustainable for many brands, according to CNBC.
Simeon Siegel, senior managing director at Guggenheim, told CNBC that proactively minimizing returns can be a meaningful driver of profitability. Siegel noted that AI-generated visuals can now be delivered to consumers in the cloud cheaply enough for brands to generate a return on the investment.
Earlier virtual try-on attempts, beginning in the 2010s, relied on static image overlays that could not simulate how a garment drapes across different body types. Researchers at the University of Illinois Urbana-Champaign have developed a video diffusion framework called Dress&Dance that produces high-resolution video clips of users modeling selected garments.
The model processes photos of clothing in real-world conditions, whether hanging on a rack or worn by another person, without requiring studio-calibrated images, according to the university. A follow-on model, Virtual Fitting Room, generates arbitrarily long try-on videos with user-controlled layering of multiple items, the university said.
Retailers Embed Try-On
Catches has developed a platform it describes as delivering mirror-like realism, building a digital twin from a user photo and incorporating the physics of fabric texture and how material interacts with a moving body. The application went live in March on luxury brand Amiri’s website. Catches projects the platform can drive a 10% increase in conversions and a 20-to-30-times return on investment for brand partners, figures the company itself provided according to CNBC.
Google expanded its AI virtual try-on feature to Australia, Canada and Japan in October and added shoe try-on to the existing apparel capability. The feature also allows users to try on the shoes virtually. Google also separately launched Doppl, which helps consumers use AI to visualize outfits by dropping a picture or screenshot.
Perplexity launched its own virtual try-on tool in November for Pro and Max subscribers, generating results in under a minute and accounting for posture, body shape and fabric drape.
Shopify has integrated startup Genlook’s AI virtual try-on app into its commerce platform, which the company says removes sizing doubts and drives higher conversion rates.
Retailers are also embedding these tools on their sites. For example, Zara rolled out a virtual try-on tool in December alongside return fees for online orders, a combination that helped protect its gross margin and reduce bracketing, the practice of ordering multiple sizes with the intent to return most of them.
ASOS recently reported a 160-basis-point reduction in its returns rate, partly from experimenting with virtual try-ons in partnership with deep-tech startup AIUTA.
The technology is still being refined, and no platform has published a definitive return rate reduction tied solely to virtual try-on. But the direction from Google, Perplexity, Shopify, Zara, and ASOS points to an industry that has decided the experiment is worth running at scale.
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
